基于动态多层次特征提取的高光谱解混方法
DOI:
CSTR:
作者:
作者单位:

沈阳理工大学

作者简介:

通讯作者:

中图分类号:

基金项目:

辽宁省属本科高校基本科研业务费专项资金资助(LJ212410144018,SYLUGXRC202258)


Hyperspectral Unmixing Based on Dynamic Multi-level Feature Extraction
Author:
Affiliation:

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    针对基于视觉Transformer进行高光谱解混时,固定大小的patch划分方式可能导致图像局部结构的坍塌和图像间语义不一致的问题,提出一种数据驱动的方法,旨在根据输入的高光谱图像自适应地调整patch划分的位置和尺度,并通过引入多尺度变形注意力机制有效捕获图像中的局部特征和空间位置信息,同时通过多层次特征提取策略,全面挖掘图像中蕴含的丰富特征。在多个真实数据集上的实验对比结果表明,所提方法能够精确估计丰度图和端元光谱,并在不同场景下,特别是在处理复杂地物光谱特征时,展现出优异的泛化能力和稳定性。

    Abstract:

    In view of the problem that the fixed-size patch division method may lead to the collapse of the local structure of the image and the inconsistency of semantics between images when performing hyperspectral unmixing based on visual Transformer, a data-driven method is proposed, which aims to adaptively adjust the position and scale of patch division according to the input hyperspectral image, and effectively capture the local features and spatial position information in the image by introducing a multi-scale deformation attention mechanism. At the same time, a multi-level feature extraction strategy is used to comprehensively mine the rich features contained in the image. Experimental comparison results on multiple real datasets show that the proposed method can accurately estimate the abundance map and endmember spectrum, and exhibit excellent generalization ability and stability in different scenarios, especially when dealing with the spectral features of complex objects.

    参考文献
    相似文献
    引证文献
引用本文
分享
相关视频

文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2025-02-20
  • 最后修改日期:2025-04-08
  • 录用日期:2025-05-08
  • 在线发布日期:
  • 出版日期:
文章二维码